A computer-implemented method for real-time transaction fraud vetting, including: receiving a transaction record including real-time transaction data and a key identifying a transacting entity; matching the key to a plurality of profiles of the transacting entity, each of the plurality of profiles including a data attribute; accessing a datapoint for each of the plurality of profiles, each datapoint representing a standard for the corresponding data attribute computed from historical transaction records of the transacting entity; assessing whether deviation of the real-time transaction data from each datapoint exceeds a corresponding threshold; incrementing a transaction risk corresponding to the transaction record for each deviation from one of the datapoints that exceeds the corresponding threshold; outputting a fraud score based at least in part on the transaction risk; and updating at least one datapoint based on the corresponding real-time transaction data to generate an updated datapoint set for the plurality of profiles.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
2. The computer-implemented method of claim 1, further comprising decrementing, via the one or more processors, the transaction risk corresponding to the transaction record for each instance in which real-time deviation of the transaction record from the one or more normal values of one of the plurality of long-term profiles does not exceed the corresponding threshold.
3. The computer-implemented method of claim 2, wherein incrementing the transaction risk based on one of the plurality of long-term profiles includes updating a corresponding real-time profile based on the transaction record and comparing the real-time profile against the long-term profile to determine the deviation.
6. The computer-implemented method of claim 5, further comprising adjusting, via the one or more processors, a second threshold of the second applied fraud model based on the incremented transaction risk from the first applied fraud model.
This invention relates to fraud detection systems that use multiple fraud models to assess transaction risk. The problem addressed is the need to dynamically adjust fraud detection thresholds based on risk assessments from different models to improve accuracy and reduce false positives. The method involves applying a first fraud model to a transaction to determine an initial transaction risk score. If the risk score exceeds a first threshold, the transaction is flagged as potentially fraudulent. Additionally, the method applies a second fraud model to the same transaction, which may use different criteria or data sources than the first model. The second model generates a second risk score, which is compared to a second threshold. If the second risk score exceeds this threshold, the transaction is further flagged or blocked. A key aspect of the invention is dynamically adjusting the second threshold of the second fraud model based on the incremented transaction risk from the first model. For example, if the first model assigns a high-risk score, the second threshold may be lowered to increase sensitivity, ensuring that the second model also flags the transaction if it detects any additional suspicious activity. This adjustment improves the system's ability to detect complex fraud patterns that may not be evident from a single model's assessment. The method ensures that fraud detection is more robust by leveraging multiple models and dynamically adapting their thresholds based on cross-model risk assessments.
9. The computer-implemented method of claim 8, wherein computing the final fraud score includes retrieving one or more user-tuned weighting adjustments and incorporating the weighting adjustments into the weighted summation.
11. The monitoring payment network server of claim 10, wherein the computer-executable instructions further cause the at least one processor to decrement the transaction risk corresponding to the transaction record for each instance in which real-time deviation of the transaction record from the one or more normal values of one of the plurality of long-term profiles does not exceed the corresponding threshold.
12. The monitoring payment network server of claim 11, wherein incrementing the transaction risk based on one of the plurality of long-term profiles includes updating a corresponding real-time profile based on the transaction record and comparing the real-time profile against the long-term profile to determine the deviation.
15. The monitoring payment network server of claim 14, wherein the computer-executable instructions further cause the at least one processor to adjust a second threshold of the second applied fraud model based on the incremented transaction risk from the first applied fraud model.
16. The monitoring payment network server of claim 14, wherein the computer-executable instructions further cause the at least one processor to increment the incremented transaction risk in real-time at least in part by implementing a velocity vector pointing to a rolling list of a long-term profile of the second applied fraud model to determine whether the second transaction record causes the rolling list to exceed a second threshold.
18. The monitoring payment network server of claim 17, wherein computing the final fraud score includes retrieving one or more user-tuned weighting adjustments and incorporating the weighting adjustments into the weighted summation.
20. The computer-readable storage media of claim 19, wherein, when executed by the at least one processor, the computer-readable instructions further cause the at least one processor to decrement the transaction risk corresponding to the transaction record for each instance in which real-time deviation of the transaction record from the one or more normal values of one of the plurality of long-term profiles does not exceed the corresponding threshold.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
April 30, 2021
October 25, 2022
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.